Scoring rules for statistical models on spheres. (English) Zbl 1463.62158
Summary: We introduce a novel class of strictly proper scoring rules for statistical models on spheres that does not require the calculation of normalizing constants. The class contains the Hyvärinen scoring rule investigated by K. V. Mardia et al. [“Score matching estimators for directional distributions”, Preprint, arXiv:1604.08470].
Software:
hgm RReferences:
| [1] | Dawid, A. P., The geometry of proper scoring rules, Ann. Inst. Statist. Math., 59, 77-93, (2007) · Zbl 1108.62009 |
| [2] | Ehm, W.; Gneiting, T., Local proper scoring rules of order two, Ann. Statist., 40, 609-637, (2012) · Zbl 1246.86013 |
| [3] | Hyvärinen, A., Estimation of non-normalized statistical models by score matching, J. Mach. Learn. Res., 6, 695-709, (2005) · Zbl 1222.62051 |
| [4] | Kume, A.; Wood, A. T.A., Saddlepoint approximations for the Bingham and Fisher-Bingham normalising constants, Biometrika, 92, 465-476, (2005) · Zbl 1094.62063 |
| [5] | Mardia, K. Kent,J., Laha, A., 2016. Score matching estimators for directional distributions. arXiv:1604.08470; Mardia, K. Kent,J., Laha, A., 2016. Score matching estimators for directional distributions. arXiv:1604.08470 |
| [6] | Mardia, K.; Jupp, P., Directional statistics, (2000), Wiley Chichester · Zbl 0935.62065 |
| [7] | Nakayama, H.; Nishiyama, K.; Noro, M.; Ohara, K.; Sei, T.; Takayama, N.; Takemura, A., Holonomic gradient descent and its application to Fisher-Bingham integral, Adv. Appl. Math., 47, 639-658, (2011) · Zbl 1226.90137 |
| [8] | Parry, M.; Dawid, A. P.; Lauritzen, S., Proper local scoring rules, Ann. Statist., 40, 561-592, (2012) · Zbl 1246.62011 |
| [9] | van der Vaart, A., Asymptotic statistics, (1998), Cambridge University Press · Zbl 0910.62001 |
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. In some cases that data have been complemented/enhanced by data from zbMATH Open. This attempts to reflect the references listed in the original paper as accurately as possible without claiming completeness or a perfect matching.
